Cochrane ‘Living’ Systematic Review on Diagnostic Accuracy of Imaging for COVID-19

Abstract

Background: The coronavirus disease 2019 (COVID-19) presents diagnostic evaluation and patient management challenges, including uncertainty regarding the role of imaging tests. This series of reviews from the suite of Cochrane ‘living systematic reviews’ aims to evaluate the accuracy of chest imaging (computed tomography (CT), X-ray and ultrasound) for diagnosis and management of individuals with suspected COVID-19. Methods: The Bern COVID-19 Living Database, Cochrane COVID-19 Register, and CDC Library were searched through 30 September 2020 (for review version 3) and 17 February 2021 (for review version 4). Diagnostic accuracy studies involving participants with suspected COVID-19 were included. Screening, data extraction, and risk of bias assessments (using the QUADAS-2 tool) were completed independently, in duplicate. Pooled accuracy estimates and 95% confidence intervals (CIs) were determined using a bivariate random effects model. Results: In the third version of the review, chest CT (41 studies, 16133 participants, 8110 (50%) cases) had a pooled sensitivity of 87.9% (95%CI 84.6-90.6) and specificity of 80.0% (74.9-84.3). Chest X-ray (9 studies, 3694 participants, 2111 (57%) cases) had a pooled sensitivity of 80.6% (69.1-88.6) and specificity of 71.5% (59.8-80.8). Ultrasound of the lungs (5 studies, 446 participants, 211 (47%) cases) had a pooled sensitivity of 86.4% (72.7-93.9) and specificity of 54.6% (35.3-72.6). Indirect comparisons showed that chest CT gave higher specificity (P=0.0052) and similar sensitivity (P=0.77) compared to ultrasound. There were no differences (P≥0.05) in accuracy between CT and X‐ray, or X‐ray and ultrasound. In the fourth version of the review, chest CT (69 studies, 28285 participants, 14342 (51%) cases) had a pooled sensitivity of 86.9% (83.6-89.6) and specificity of 78.3% (73.7-82.3). Chest X‐ray (17 studies, 8530 participants, 5303 (62%) cases) had a pooled of sensitivity=73.1% (64.1-80.5) and specificity of 73.3% (61.9-82.2). Ultrasound of the lungs (15 studies, 2410 participants, 1158 (48%) cases) had a pooled sensitivity of 88.9% (84.9-92.0) and specificity of 72.2% (58.8-82.5). Indirect comparisons showed that chest CT and ultrasound had similar sensitivities (P=0.42), and each gave higher sensitivities than X-ray (P=0.0003 and P=0.001, respectively). All modalities had similar specificities (P≥0.05). Conclusion: The most recent evidence indicates that both chest CT and ultrasound of the lungs are sensitive and moderately specific for diagnosing individuals with suspected COVID-19, while chest X-ray is moderately sensitive and moderately specific. Chest CT and ultrasound may be useful for ruling out COVID‐19, but not for distinguishing COVID-19 from other illnesses. Research assessing the prognostic value of imaging for predicting morbidity and mortality in individuals with COVID-19 is underway and will also be published in the suite of Cochrane ‘living' systematic reviews.